Helper Functions

General utility functions for Logical Neural Networks.

jlnn.utils.helpers.is_precise(interval: Array, epsilon: float = 1e-05) → bool[source]

Checks if a truth interval has collapsed into a single point (L ≈ U).

Parameters:

  • interval – Truth interval tensor [L, U].

  • epsilon – Maximum allowed difference between bounds.

Returns:

True if the uncertainty is within epsilon, False otherwise.

jlnn.utils.helpers.scalar_to_interval(x: Array) → Array[source]

Converts standard [0, 1] probability scalars into precise [L, U] intervals.

This is used to ground the JLNN model with data from classical datasets where truth values are known exactly (L = U = x).

Parameters:

x (jnp.ndarray) – Tensor of scalar truth values.

Returns:

Tensor of intervals with shape (*x.shape, 2).

Return type:

jnp.ndarray

Helper functions for transforming data between standard formats and the JLNN interval representation.

  • scalar_to_interval: Converts classical probabilities in the range \([0, 1]\) to precise intervals \([x, x]\). This is crucial when initializing facts from fixed datasets.

  • is_precise: Checks whether the interval has already “collapsed” into a single point (zero uncertainty).